Screening and control of methicillin-resistant Staphylococcus aureus in 186 intensive care units: different situations and individual solutions

Detailed methods of the modules of the German Nosocomial Infection Surveillance System for surveillance of nosocomial infections and multidrug-resistant bacteria in ICUs have been described previously [5, 15]. In short, the surveillance module for multidrug-resistant bacteria (MDR-KISS) collects, among other multidrug-resistant bacteria data, data on all admitted patients with MRSA, including colonized patients and patients with infections not fulfilling the definitions of nosocomial infections of the Centers for Disease Control and Prevention. Surveillance of MRSA started in 2003, and 240 ICUs regularly submitted data on MRSA cases at the beginning of this study in 2007. Surveillance is performed for all patients admitted to these ICUs; however, patient-based data are collected only for patients with a culture that was positive for MRSA and that was recovered during the ICU stay or for patients known to be carriers on admission. Cases are differentiated between imported and ICU-acquired on the basis of a temporal definition with a 48-hour interval between admission and detection of MRSA as a cutoff. The following data are collected for every patient carrying MRSA in participating ICUs: dates of admission, discharge, and first isolation of MRSA; import of MRSA (defined as known carriage or detection of MRSA in any clinical or surveillance culture within 48 hours of ICU admission) or acquisition during the ICU stay (defined as detection within more than 48 hours of ICU admission); presence of colonization or presence of an infection that requires treatment during the ICU stay; and, in case of infection, the site of infection. For the denominator, the total number of patient-days (pd) is determined without excluding the patient-days of MRSA-positive patients.

The data are collected and stored anonymously and according to national guidelines for data protection. All data collected by MDR-KISS and included in this study are obtained during routine surveillance required by the German Protection against Infection Act (Infektionsschutzgesetz) [16]. In paragraph 23 of this law, hospitals are obliged to continuously collect and analyze data on nosocomial infections and resistant pathogens. No additional data not covered by this law are used for this study. Ethical approval and informed consent are, thus, not required. Data are collected by trained medical or infection control personnel in individual ICUs and are submitted via a web-based surveillance portal to the central database of the national reference center. MRSA isolates of submitted cases are not collected for further analysis such as molecular typing. Mean and median IDs of MRSA cases per 1,000 pd with the interquartile ranges are calculated for all MRSA cases and the subcategories of imported and ICU-acquired cases from the pooled data of all ICUs as a reference for comparison with local MRSA IDs.

1. MRSA case: any patient who is carrying MRSA and who is admitted to participating ICUs, including patients with infection and colonization as well as imported and ICU-acquired cases.

2. Imported MRSA case: a patient with known carriage of MRSA on admission or detection of MRSA in any clinical or surveillance culture within 48 hours of admission.

3. ICU-acquired MRSA case: a patient with detection of MRSA in any clinical or surveillance culture within more than 48 hours of admission.

In 2008, a detailed questionnaire with 75 questions regarding ICU structure (for example, number of beds, single rooms, and health-care staff), microbiological diagnostics (for example, routine urine and respiratory surveillance cultures), MRSA screening procedures (for example, polymerase chain reaction-based or culture-based screening and patient population screened), MRSA control measures (for example, single-room isolation, pre-emptive isolation, and decolonization protocols), infection control education, and implementation of surveillance was sent to participating ICUs. Questions were directed at control practices performed in 2007. All ICUs participating in the MDR-KISS surveillance network were asked to complete the questionnaire; ICUs were not chosen prospectively for performing or not performing screening. Answers to the electronic questionnaire were submitted online to the national reference center.

To describe the use of combinations or bundles of control measures, a tree diagram, including the three key control parameters of screening policy, isolation of MRSA carriers, and decolonization, was constructed. For this diagram, three categories of screening (universal admission screening, targeted screening, and no screening) and three categories of isolation practices were distinguished. The categories for isolation practices were strict single-room isolation defined as isolation of all MRSA carriers in single rooms; single-room isolation or contact precautions defined as isolation of MRSA carriers in single rooms if possible (but if no single room was available, contact precautions in larger rooms were performed); contact precautions defined as performance of contact precautions for MRSA carriers in larger rooms and no attempt of single-room isolation. Median IDs of MRSA cases per 1,000 pd with interquartile ranges were calculated for nine subgroups of ICUs implementing different screening and isolation strategies.

Generalized linear models (GLMs) were used to estimate the association between parameters derived from the questionnaire and the number of all MRSA cases and the number of ICU-acquired MRSA cases, respectively. In the analysis, we considered the three key control parameters described above and their interactions, pre-emptive isolation, and the structural parameters of hospital type and size and ICU type and size, short admissions of less than 48 hours of less than one third of admitted patients, the percentage of ventilator beds, the number or percentage of single rooms, and the nurse-to-patient ratio. The nurse-to-patient ratio was calculated as the number of nurses per shift per patient. For employed nurses in 2007, a 40-hour week, not including sick or vacation days, was considered.

Additionally, to adjust better for the severity of disease, the length of stay and device use (use of central venous catheters and ventilation per 100 pd) were considered as potential confounding parameters in the model. We used negative binomial distribution in the model instead of Poisson distribution because the variance exceeds the mean and we observed overdispersion for the number of MRSA cases. The log number of patient-days was treated as an offset in the model. A stepwise forward approach was applied for the multivariate analysis. Selection criteria were the highest chi-square value and a P value of less than 0.05 in the type III score statistic. The Akaike information criterion was used as goodness-of-fit measure in the GLM. P values of less than 0.05 were considered significant. All analyses were performed with SPSS (IBM SPSS Statistics; IBM Corporation, Armonk, NY, USA) and SAS (SAS Institute Inc., Cary, NC, USA).

Two hundred forty ICU-KISS ICUs with at least 1 month of MDR-KISS data in the period from 2003 to 2007 were asked to complete the questionnaire, and 186 ICUs responded (response rate of 77.5%). There was no significant difference by hospital type and size or ICU type and size between responding and non-responding ICUs. The 186 responding ICUs admitted 302,524 patients with 1,075,364 pd and 430,685 ventilator-days from 2007 to 2008. Four thousand nine hundred thirty-five MRSA cases occurred in these ICUs during the period of analysis; of these, 3,928 (79.6%) were imported and 1,007 MRSA cases (20.4%) were ICU-acquired according to the definitions. Calculated per ICU, this translates to a median of 9 MRSA cases (7 imported and 2 acquired) per year with a range from 0 to 163 yearly MRSA cases. Structural characteristics as well as device use and the IDs of MRSA cases in the 186 study ICUs are shown in Table 1.

The results of the questionnaire for the most relevant MRSA screening and control measures are shown in Table 2. Individual MRSA control measures were implemented in several different combinations. Taking into account the three key control parameters of screening policy, isolation of identified carriers, and decolonization resulted in 17 different combinations (Figure 1). The most common combination of control measures used by ICUs was targeted screening, strict single-room isolation, and decolonization with mupirocin (Figure 1). The MRSA IDs of the 186 ICUs stratified by the implemented screening and isolation measures are shown in Table 3. The MRSA IDs of the different groups of ICUs varied significantly for all MRSA cases and imported MRSA cases but not for ICU-acquired MRSA cases (Table 3). A comparison of ICUs performing universal admission screening with ICUs performing targeted or no screening revealed that ICUs with universal admission screening had a significantly higher number of ICU and hospital beds, longer length of stay, and higher device use than ICUs with other screening policies (data not shown).

The total number of MRSA cases per ICU was independently associated with the MRSA screening policy and the size of the hospital in the GLM (Table 4). Analyses of factors associated with ICU-acquired MRSA cases in ICU subgroups stratified according to screening strategies show an independent association of the type of ICU and the incidence of imported MRSA cases in the model with 51 ICUs with universal admission screening (Table 5); an independent association of the type of ICU, the incidence of imported MRSA cases, and the length of stay in the model with 91 ICUs with targeted screening (Table 6); and an independent association of the incidence of imported MRSA cases in the model with 44 ICUs without screening (Table 7) with ICU-acquired MRSA cases, respectively.

Although MRSA was shown to be endemic in ICUs of all German regions [5], the MRSA IDs in the analyzed ICUs are not uniform. The MRSA burden in ICUs ranges from ICUs with no cases within the two-year period of analysis to 30.1 MRSA cases per 1,000 pd. Previous studies have shown that the benefit to be expected from active surveillance cultures is dependent on the local reservoir of MRSA carriers [21]. The variation between the MRSA IDs of ICUs in this study suggests that the extent of control efforts needed and their potential benefit for patients will differ considerably between ICUs.

A comparison of the MRSA IDs of the ICUs in our sample stratified according to screening strategies shows that ICUs performing universal admission screening have significantly higher MRSA IDs than ICUs with targeted or no MRSA screening. However, it remains unclear to which extent the high MRSA IDs in ICUs with universal screening programs are the cause (implementation of screening because of a perceived MRSA problem) or the effect (better detection of MRSA carriers) of MRSA screening. A comparison of structure and process parameters shows that ICUs performing universal screening are high-risk units for MRSA with higher use of invasive devices, longer length of stay, and a higher number of ICU and hospital beds than ICUs without universal screening programs. In addition, the regression analysis shows an independent association of the size of the hospital and screening policies with MRSA cases. The influence of the structural parameter and the distortion caused by screening suggests that the total number of MRSA cases is not a suitable parameter for benchmarking of ICUs.

Substantial variation in MRSA control measures between European countries has been described [22]. Our study shows that the implemented control measures also differ considerably between ICUs within one country. The evidence for the effectiveness of individual components of MRSA control such as screening, isolation, and decolonization has been described as weak [23]. In addition, the wide range of implemented combinations of control measures in association with differences in hospital structure, patient population, and imported MRSA incidence might explain the difficulty to evaluate the benefit of MRSA screening programs. The variety of different MRSA situations in the 186 ICUs indicates that studies of specific MRSA control strategies performed in a single ICU will have very limited general applicability.

Factors associated with ICU-acquired MRSA cases in the GLMs analyzing three subgroups of ICUs with different screening strategies include the type of ICU, the imported MRSA incidence, and the length of ICU stay. The impact of these factors has been shown before [24, 25] and is also further investigated in a separate analysis of a subgroup of ICUs from our sample [26]. The models show no association between a single MRSA screening or control measure and ICU-acquired MRSA cases; however, screening and other control measures are unlikely to be successful as isolated measures. A decision-analytical model has shown that the combination of active surveillance and decolonization was more effective than active surveillance alone [27], but the importance of individual components of MRSA control programs and the minimum effective measures needed for MRSA control are not known [28]. We therefore aimed to assess the interactions of control measures; however, after stratification of ICUs into subgroups according to implemented screening strategies to minimize MRSA ID distortion due to screening, no combination of isolation and decolonization measures was found to be associated with ICU-acquired MRSA cases.

The incidence of imported MRSA cases was associated with ICU-acquired cases independently of the implemented screening strategy and the effort made to identify MRSA carriers. This result indicates that the incidence of imported MRSA cases could provide a hint for the extent of control measures needed in ICUs.

In spite of including combinations of relevant control strategies and adjusting for the imported incidence of MRSA and proxy measures for the admitted patient population and disease severity, our model will fall short of capturing the complexity of MRSA control in ICUs. Further limitations are that the performed association analysis does not allow any conclusions about causation and that it is based on survey results that have not been validated by observation. We did not collect data on other important MRSA control variables such as compliance with hand hygiene and used a temporal definition to differentiate between import and acquisition of MRSA, and this might lead to a misclassification of a number of cases. Molecular typing was not performed, and the number of MRSA cases attributable to epidemic MRSA strains, community-acquired MRSA, and livestock-associated MRSA in participating ICUs remains unclear; however, a low percentage of community-acquired and livestock-associated MRSA in MRSA isolates from German hospitals has been reported [29].